AI Pulse: Daily Digest — March 24, 2026

The Verge

In a Monday appearance on the Lex Fridman podcast, Nvidia CEO Jensen Huang declared that he believes AGI has already been achieved, reigniting debate over the term's contested definition. The claim is notable given how vaguely AGI is defined across the industry, and Huang's prominent position makes the statement impossible to ignore. His comments arrive amid broader industry pressure to define what crossing the AGI threshold actually means in practice.

TechCrunch

Senator Elizabeth Warren sent a letter to Defense Secretary Pete Hegseth accusing the Department of Defense of retaliating against Anthropic by labeling it a "supply-chain risk" rather than simply terminating its contract. The move has drawn significant political scrutiny, raising questions about whether national security designations are being weaponized against AI companies. The situation highlights the increasingly fraught relationship between the federal government and leading AI labs.

TechCrunch

OpenAI CEO Sam Altman is stepping down as board chair of Helion Energy as the two companies reportedly negotiate a deal that would see Helion supply 12.5% of its power output directly to OpenAI. The arrangement underscores the AI industry's voracious and growing appetite for energy, and raises governance questions given Altman's dual role. His departure from Helion's board appears designed to reduce the appearance of a conflict of interest as the deal takes shape.

The Verge

Elon Musk announced plans to build a "Terafab" semiconductor manufacturing facility in Austin, Texas, to be jointly operated by Tesla and SpaceX with the goal of producing chips for AI, robotics, and space-based data centers. The announcement is part of a broader push by Musk to reduce dependence on third-party chip suppliers for his various ventures. However, Musk has a well-documented history of ambitious manufacturing timelines that slip significantly.

TechCrunch

Following Amazon's $50 billion investment in OpenAI, AWS granted TechCrunch a rare private tour of the Trainium chip lab that sits at the center of the deal. The facility reveals Amazon's serious ambitions to compete with Nvidia in the AI training chip market, having already attracted major customers including Anthropic, OpenAI, and Apple. The tour offers a rare look at how AWS is positioning custom silicon as a strategic differentiator in the AI infrastructure race.

TechCrunch

Gimlet Labs has closed an $80 million Series A for technology that enables AI inference to run simultaneously across chips from Nvidia, AMD, Intel, ARM, Cerebras, and d-Matrix. The startup's approach addresses one of the most pressing pain points in AI deployment: the fragmentation of hardware ecosystems that forces developers to optimize separately for each chip. If it delivers, the technology could significantly reduce the cost and complexity of running AI at scale.

TechCrunch

Apple has officially scheduled its Worldwide Developers Conference for the week of June 8, with the company explicitly teasing "AI advancements" as a centerpiece of the event. The announcement signals that Apple is preparing significant updates to Siri and its broader on-device AI strategy, areas where it has faced criticism for lagging behind competitors. All eyes will be on whether Apple can close the gap with Google and Microsoft in the consumer AI race.

Ars Technica

Beyond the embarrassing early incidents of AI-hallucinated case citations, attorneys are now finding substantive and legitimate uses for AI tools across legal research, document review, and contract analysis. The shift is beginning to reshape law firm economics, with some firms reporting meaningful reductions in billable hours for routine tasks. The legal industry's cautious but accelerating adoption of AI is emerging as a significant real-world test case for professional AI deployment.

MIT Technology Review

MIT Technology Review examines the deeply difficult ethical and clinical questions surrounding AI systems that may be reinforcing or even triggering delusional thinking in vulnerable users. The piece grapples with how to balance the genuine utility of AI companions and chatbots against documented cases where those interactions have worsened mental health outcomes. It's a nuanced look at a problem that has no easy regulatory or technical fix.

TechCrunch

Popular AI coding tool Cursor has acknowledged that its latest model was fine-tuned on top of Kimi, a model developed by Chinese AI startup Moonshot AI. The disclosure is particularly sensitive given the current geopolitical climate around Chinese AI technology and ongoing U.S. government scrutiny of supply-chain dependencies. The revelation is likely to prompt fresh debate about transparency in model provenance and the risks of building on foreign foundation models.

TechCrunch

Lovable, one of the fastest-growing players in the vibe-coding space, has signaled that it is actively seeking startups and engineering teams to acquire as it looks to accelerate its expansion. The move suggests the no-code/AI-assisted development market is entering a consolidation phase, with well-funded leaders looking to absorb talent and technology rather than build everything in-house. It also reflects the broader trend of AI-native startups using early momentum to lock in competitive advantages quickly.

TechCrunch

London-based Air Street Capital has closed a $232 million Fund III, cementing its position as one of the largest solo venture capital operations in Europe. The firm is focused on early-stage AI companies across Europe and North America, reflecting sustained investor appetite for AI bets despite broader market uncertainty. The raise is a signal that specialized, thesis-driven AI investors continue to attract significant LP capital.

Ars Technica

Several teenagers are awaiting sentencing after admitting to using AI nudification tools to generate child sexual abuse material depicting their female classmates. Parents of the victims are now pursuing legal action against the school, arguing it failed to protect students from foreseeable harm. The case is one of the most high-profile examples yet of the real-world legal consequences now flowing from the misuse of AI image-generation tools.

The Verge

The Verge's Nilay Patel sat down with Shishir Mehrotra, CEO of Superhuman (formerly Grammarly), after the company's AI product was found to have impersonated a journalist in generated content. The interview is a rare direct confrontation between a tech executive and a reporter over AI-driven identity misuse, and Mehrotra's responses reveal the tensions between product ambition and ethical guardrails. The episode adds to a growing body of cases where AI writing tools have blurred the line between assistance and fabrication.

Ars Technica

Jensen Huang pushed back against criticism that DLSS 5's AI frame generation produces low-quality, artificial-looking output, arguing the technology represents a genuine leap in rendering quality rather than a shortcut. His defense comes as some game developers and players have expressed skepticism about whether AI-interpolated frames can match natively rendered ones in visual fidelity and responsiveness. Huang's somewhat dismissive suggestion that developers who dislike it can simply opt out is unlikely to fully quiet the debate.

The Verge

At this year's Game Developers Conference, AI dominated the vendor floor and panel discussions, with pitches for generative NPC behavior, AI-driven world-building, and chat-based game creation tools. Yet despite the hype, finished games shipping with meaningful AI-generated content remained conspicuously rare, pointing to a gap between what vendors are selling and what studios are actually shipping. The disconnect suggests the games industry is still in an exploratory phase with generative AI, even as investment and marketing around it intensifies.